The Evolution of Predictive Analytics
Predictive analytics has been a useful insurance industry tool for many years, but the practice is evolving. We cover the past, present, and future of predictive analytics.
February 7, 2023
Using predictive analytics in insurance claims can help identify risks; it is a wrench within the toolbox of workers’ compensation risk management and is continuously progressing.
“Predictive analytics has evolved significantly,” said Bill Wilkins, Vice President and Chief Risk & Analytics Officer at Safety National. “In the 1960s, Allstate started using data similarly to what others are doing today culminating in the Allstate Research and Planning Center, where in 1982, they started investigating the theory that red cars are more ticketed. The entire purpose of this project was to gather the existing computer system information and make the connections. That’s exactly what predictive analytics is all about — making the right connections that provide the most powerful information for the future.”
The use of predictive analytics goes back further than in the 1960s. It was utilized in the 1940s, during World War II, by the U.S. Navy to determine the safest route for cargo ships by attempting to locate enemy U-boats. The military did not have the advanced technology we do now, so most predictive analytics came from gathering information by hand and using mathematical techniques.
Decades ago, a very limited number of people were practicing predictive analytics. Today, we have companies like DataRobot that let users put predictive analytics in an AI platform and realize the capabilities of this technique. We have now made the computer able to perform calculations once reserved for the skills of mathematicians. While strikingly advanced, it has drawbacks, including the quality and inherent bias of the data used.
The current challenge is determining the best — and unbiased — data to use for predictive analytics. Biased data is, and will continue to be, the thorn in the side of predictive analytics users. A great data scientist must understand that there can and often will be data biases.
In data analytics, it is vital not to take anything at face value. If you see a trend, investigate why that trend is there. The public has become more skeptical of data due to frequent mistakes, slowing down the acceptance and progress of predictive analytics, so professionals must perform due diligence in trends investigation.
The future of predictive analytics lies in education. If the insurance industry cannot provide the reasoning behind machine-based data analysis, this profession will not continue to succeed. There are significant concerns that users in the future will be consuming the data and repurposing it, but not investigating potential bias. When this data is used to make critical decisions, it should always be questioned and triple-checked for accuracy.
Predictive analytics can be one of the most valuable tools for injury and accident prevention. For example, Safety National had worked with a large hotel chain experiencing an alarming amount of shoulder injuries among its housekeeping staff. It quickly became a very noticeably odd trend. Working with the client and their risk management group, Safety National discovered that the hotel purchased new beds, which were much heavier than the previously used beds. The staff responsible for making the beds were injuring their shoulders in the process. The solution was a tool that wedged between the bed and bed frame to lift the bed, making it easier to change the sheets. Because this solution was implemented, reported shoulder injuries — and subsequent workers’ compensation claims — decreased significantly.
Predictive analytics has that capability and more if it can be transparent and unbiased.